Time for a neonatal–specific consensus definition for sepsis

Abstract

Objective

To review the accuracy of the pediatric consensus definition of sepsis in term neonates and to determine the definition of neonatal sepsis used.

Study selection

The review focused primarily on pediatric literature relevant to the topic of interest.

Conclusions

Neonatal sepsis is variably defined based on a number of clinical and laboratory criteria that make the study of this common and devastating condition very difficult. Diagnostic challenges and uncertain disease epidemiology necessarily result from a variable definition of disease. In 2005, intensivists caring for children recognized that as new drugs became available, children would be increasingly studied and thus, pediatric-specific consensus definitions were needed. Pediatric sepsis criteria are not accurate for term neonates and have not been examined in preterm neonates for whom the developmental stage influences aberrations associated with host immune response. Thus, specific consensus definitions for both term and preterm neonates are needed. Such definitions are critical for the interpretation of observational studies, future training of scientists and practitioners, and implementation of clinical trials in neonates.

Statement of the problem

Neonatal sepsis results in death or major disability for 39% of those affected even with timely antimicrobial treatment¹. The incidence of severe sepsis in newborns doubled (from 4.5 to 9.7 cases per 1,000 births) between 1995 and 2005². The frequency of sepsis during the birth hospitalization varies inversely with gestational age at birth and may reach 60% in the most immature infants³. The short-term economic burden of caring for and hospitalizing these infected infants is staggering and is estimated at approximately $700 million in the US⁴. Neither the treatment of neonatal sepsis, nor the neurodevelopmental outcomes in surviving infants has changed significantly over the last thirty years despite multiple failed attempts to reduce the burden of infection^5,6. These disappointments have occurred in the context of tremendous advances in other areas of newborn care including nutrition, management of respiratory distress and pulmonary hypertension, and therapeutic cooling following hypoxic-ischemic encephalopathy.

Adult and pediatric intensivists currently use consensus definitions for sepsis for goal-based therapeutic interventions^7–10. These definitions are critical to facilitate epidemiologic studies, to accurately determine disease incidence, to select patients for clinical trials, to improve training, and ultimately, to improve the delivery of care. The pediatric consensus definition for sepsis, established in 2005, was intended for all children (<18 years old) and including term (≥37 weeks completed gestation) neonates (Supplemental Table 1)⁷. Preterm neonates (<37 weeks completed gestation) were specifically excluded from the pediatric consensus definitions and neonatal-perinatal subspecialists were not represented among the pediatric consensus experts. To investigate whether the pediatric consensus definitions for systemic inflammatory response syndrome (SIRS) and sepsis applied to term infants, Hofer et al retrospectively examined 476 term neonates and found that the consensus definitions applied to only 53% of cases of culture-positive early-onset sepsis (EOS)¹¹. The authors determined that the sensitivity of more comprehensive clinical and laboratory criteria to define SIRS in the setting of culture-proven EOS (n = 30) was 20% for hypothermia or fever, 43% for abnormal WBC and neutrophil indices, 87% for respiratory signs, and 33% for cardiovascular signs. To date, the accuracy of the pediatric consensus definitions has not been assessed in preterm infants, nor have consensus definitions been developed or tested in this unique developmentally immature population.

How has sepsis been defined in neonates?

There is remarkable heterogeneity among studies regarding the definition of neonatal sepsis (Supplemental Table 1). For example, in 12/42 (29%) selected studies/guidelines single or combinations of laboratory tests were incorporated into the definition of sepsis and included C-reactive protein (CRP) (n=5)^12–16, total WBC (n=4)^7,12,17,18, metabolic acidosis (n=3)^12,18,19, unspecified lab studies (n=2)^20,21, I/T ratio (n=3)^7,13,18, neutropenia (n=1)¹³, abnormal fibrinogen (n=1)¹², thrombocytopenia (n=1)¹², hyperglycemia (n=1)¹⁹ and hypoglycemia (n=1)²². In many cases, there was additional variability among lab results defined as abnormal. Clinical findings were integrated in 26/42 (62%) of the selected studies/guidelines and included unspecified signs of sepsis (n=16)^{14,15,20,21,23–34}, cardiovascular signs [tachycardia/bradycardia, hypotension, poor perfusion (n=12)]^{7,12,13,17–19,22,30,32,35,36}, respiratory signs [apnea, cyanosis, tachypnea, need for ventilator, increased oxygen requirement (n=9)]^{1,7,12,13,17–19,22,35}, abnormal temperature [fever or hypothermia (n=7)]^{7,12,17–19,22,35}, central nervous system signs [lethargy, hypotonia, seizure (n=2)]^12,22, and feeding problems (n=1¹²). In some reports, neonatal sepsis was defined by the duration of antimicrobial treatment (at least 5 or more days)^6,37–39. Thus, there is an unmet need for the development of a consensus definition for sepsis in both preterm and term neonates to be used for future clinical studies.

The pediatric consensus definition of sepsis is SIRS in the presence of or as a result of suspected or proven infection. SIRS requires either 1) abnormal white count [total white blood cell (WBC) increased or decreased for age -or- >10% immature neutrophils] or 2) abnormal core temperature (>38.5° or <36°C)⁷. Because abnormal WBC indices or an abnormal core temperature are required for the definition of SIRS, we will focus on the predictive accuracy of these tests in term and preterm neonates.

Total leukocyte counts and neutrophil indices are inaccurate to define sepsis in term or preterm neonates

The complete blood count (CBC) with differential is commonly ordered with a sepsis evaluation in the NICU. Many factors can affect the WBC and differential counts including maternal hypertension, the method of delivery, the infant’s sex, age in hours, and the method of blood sampling⁴⁰. The impact of gestational age on the range of normal WBC indices must also be taken into account. Following specific exclusion of infants with sepsis or suspected sepsis, Schmutz et al studied the range of ANC in >30,000 CBCs obtained within 72 hours of birth from infants of 23 to 42 weeks completed gestation including 852 infants with <28 weeks gestation⁴¹. The upper and lower limits for the ANC from birth to the 3^rd day of life were 1500 to 41000/µL; from the 3^rd day until the 10^th day of life counts ranged from 1100 to 15300/µL.

Hornik et al retrospectively examined the utility of the CBC and WBC indices (total leukocyte counts, immature to total neutrophil (I/T) ratio and absolute neutrophil counts [ANC]) from >200,000 infants (including both preterm and term infants), and concluded no CBC parameter had sufficient utility for identifying infants with early-onset or late-onset sepsis^42,43. The area under receiver operating characteristics curves (AUC) was ≤0.668 for all individual indices including WBC, I/T, ANC, and platelet counts highlighting the limited utility of the CBC and in particular the WBC to identify infants with early-onset sepsis. A significant proportion of infants (14.4%) with LOS had normal CBC indices (WBC count 5000/mm3–19,000/mm3, ANC ≥1500/mm3, I/T ratio <0.2)⁴². In a study investigating the diagnostic value of waiting a period of time to obtain a CBC after birth to identify near-term (34–36 weeks completed gestation) and term infants with blood culture-positive sepsis Newman concluded, “…even when the CBC is optimally interpreted, decisions about antibiotic treatment should remain highly dependent on maternal risk factors and newborn symptoms of infection”⁴⁰.

While pediatricians/neonatologists often reply upon a “shift to the left” to identify infected infants, the identification of band forms is problematic. Van der Meer et al examined the range of variance for band form neutrophils by surveying >750 technician responses from >100 laboratories in Europe⁴⁴. All observers received identical slides with 100 Giemsa-stained cells from a patient with sepsis. They reported wide intra- and inter-laboratory variation in interpretations for neutrophils (15–72%, SD 11%) and for band forms (4–64%, SD 11%). This study suggests the enormous variability of interpretation of the band count and thus the I/T ratio may limit use of the I/T ratio as a definition criterion.

Core temperature is inaccurate to define sepsis in term or preterm neonates

The use of an abnormal core temperature as a criterion for sepsis has significant limitations in preterm infants. Preterm neonates manifest a baseline lack of temperature maintenance due to physiologic immaturity that requires servo-controlled incubators. Thus, hypothermia may be secondary to inadequate provision of heat rather than sepsis. Among 395 symptomatic preterm infants with blood culture-positive late-onset sepsis (LOS), only 10.8% had temperature instability that occurred 24 hours before or after the first positive blood culture (positive predictive value 10.2%)³⁴. Neonates, particularly preterm neonates, rarely manifest fever greater than 38°C even with sepsis and septic shock, and are more likely to manifest temperature instability or hypothermia. Thus, an abnormal core temperature is a difficult criterion to interpret in the setting of neonatal SIRS/sepsis. Voora et al showed that among 10,092 term newborns, 1% (n=100) developed an axillary or rectal temperature ≥37.8C within 4 days of birth and only 10% of those with a temperature ≥37.8°C (n=10) had a positive blood culture⁴⁵.

Interpretation of blood culture results in defining sepsis in term or preterm neonates (Type I and Type II errors)

When a potential pathogen grows from a blood culture, it is important to distinguish true infection from a potential contaminant (type I error). As many as 62% of extremely low birth weight infants who survive >12 hours after birth have a positive blood culture during their hospitalization³. Coagulase-negative Staphylococci (CoNS) represents greater than 50% of isolated bacteria from blood cultures of preterm infants in the United States and many centers worldwide^16,46,47. Management of a single blood culture positive for CoNS varies widely, from careful observation to aggressive treatment, which implies some clinicians may view it as a contaminant⁴⁸. The rate of false positive blood cultures (of which >67% are due to CoNS) is directly proportional to the age of the child and is as high as 17% in infants <12 weeks of age⁴⁹.

Although a positive culture is often considered the “gold standard” definition of infection, culture-negative, clinical sepsis is considered a real entity in all age groups. Furthermore, a pathogen may be identified in as little as 36–51% of cases of sepsis in adults^50,51 even though sepsis is defined as SIRS in response to an inciting infection. A similar rate of culture-negative sepsis is seen in pediatric patients even in the setting of shock¹⁸. In a study of newborns with unequivocal infection documented at autopsy, premortem blood cultures were negative in 14% of cases⁵². In another study of 92 neonates ≥34 weeks with culture-positive bacterial meningitis, 35 (38%) had negative blood cultures⁵³. When blood and other sterile site cultures are negative, but the infant manifests signs concerning for infection they are considered to have clinical sepsis. In neonates, this clinical scenario is far more common than blood-culture positive sepsis⁵⁴, and represented nearly 60% of the subjects enrolled in the recent International Neonatal Immunotherapy Study (INIS) that examined treatment of neonatal sepsis with intravenous immune globulin¹. Furthermore, neonates often have low colony count bacteremia and when an insufficient volume of blood is obtained for culture, the blood culture may be falsely negative (type II error)⁵⁵. Finally, the possibility of a non-bacterial cause of infection must also be considered. Fungal and viral infections may also generate a systemic inflammatory response and sepsis. There is increasing evidence for novel viral pathogens associated with sepsis-like syndrome in preterm infants (e.g., echovirus, enterovirus, parechovirus, coxsackie, adenovirus, parainfluenza, rhinovirus, coronavirus)^56–59.

How should we move forward?

We must first establish a consensus definition of neonatal sepsis to refine future observational studies and clinical trial designs. The limited diagnostic accuracy of common laboratory tests (including the WBC indices and acute phase reactants) for diagnosing neonatal sepsis is well described (Supplemental table 2) and has been reviewed in detail elsewhere^60,61. Improved diagnostic accuracy may come from novel microscale-based molecular methods including 16sRNA DNA PCR, genome-wide expression profiling, flow cytometry-based blood analysis, and proteomic approaches. For example, by first examining patients that met the consensus definition for pediatric septic shock, Wong et al used genome-wide expression profiling on whole blood to demonstrate multiple subclinical molecular endotypes⁶². Utilization of these robust methods within a definitive disease framework is highly likely to reveal markers that facilitate early identification of patients at high risk of poor outcomes and thus permit stratification risk for future clinical trials. However, in the absence of a consensus definition for sepsis, sensitive molecular diagnostic tests derived from subjects in observational studies where multiple sepsis definitions are used (Supplemental table 1) will only further amplify the heterogeneity of the populations and would thus be at risk of demonstrating poor diagnostic accuracy between clinical cohorts.

To complement classic and modern molecular laboratory testing, defining neonatal sepsis should include a severity of illness description. The PIRO model (predisposition, insult/infection, response, organ dysfunction⁶³) is one example of a severity of illness model that has been used to study sepsis in adults and identify patients at high risk of mortality^64,65. Examples of predisposition factors for sepsis in neonates would include maternal risk factors (e.g., preterm premature rupture of membranes or vaginal colonization with group B streptococcus⁶⁶) and neonatal risk factors (prematurity or need for invasive monitoring). Infection factors might include the type and antimicrobial resistance pattern of the pathogen (bacteria, virus, or fungus-if identified) and the site of infection (e.g. blood, urine, CSF, trachea, multi-site or disseminated). Host response factors might encompass the inflammatory response (longitudinal plasma cytokines and acute phase reactants), severity and duration of clinical signs (e.g., tachypnea, increased oxygen requirement, apnea, tachycardia) and laboratory data such as neutropenia. Host response assessments may permit identification of temporal stages of sepsis pathology amenable to different treatment approaches. Organ dysfunction factors might include the requirement for inotropes, intubation for respiratory failure, oliguria, or coagulopathy. Importantly, no matter what criteria constitute the initial consensus definition for sepsis, revision and refinement of the definition should be expected to occur as new data is acquired.

Potential limitations of a consensus definition of sepsis

The present consensus definitions for sepsis in children and adults are threshold-based and thus static. A static definition of sepsis has potential advantages, but also has inherent limitations including the inability to incorporate change in status over time. For example, a patient may meet criteria (laboratory or clinical) for the definition of sepsis at one point in their disease process but not at another. Thus, a static definition of sepsis will likely be associated with limitations in diagnostic accuracy because sepsis is a dynamic, complex and heterogeneous condition⁶⁷. This challenging feature of sepsis, in addition to the biology/cutoff value used for biomarkers and variability of sepsis definitions between studies, likely contributes to the frequently limited positive predictive value but high negative predictive value seen for clinical signs and common laboratory tests used in neonates (Supplemental table 2). Although important for diagnostic accuracy, time is a difficult variable to integrate into the static decision to enroll a patient in a prospective interventional trial. Therefore, the variable of time (e.g. time of clinical presentation/biomarker measurement, duration of clinical and laboratory signs) should be collected in diagnostic studies and RCTs. Another potential limitation of a consensus definition is the background of the consensus members. A homogeneous consensus (e.g. all newborn critical care providers) is at risk of reflecting the bias of common training. Therefore, any expert consensus should have diversity among its members. Lastly, the consensus should only defer to expert opinion when data is not available. This last point illustrates the need to collect data for diagnostic and interventional studies in order to refine the criteria that constitute the definition of neonatal sepsis. Any useful definition would need to address these points and would need validation with outcomes especially in preterm neonates. A consensus definition is not meant to be the last word on how to define sepsis. Rather, the establishment of an accepted definition to be used by clinical researchers is but the first of many steps towards improving outcomes for neonates with sepsis. We have shown the present definition of neonatal sepsis is both static and variable; the combination of which we believe is entirely inadequate to allow progress.

Summary

Sepsis is a common and devastating problem in the NICU. Diagnostic and clinical interventional trials focused on neonatal sepsis are significantly hampered by use of a widely diverse definition for this disease. The first step toward addressing the poor outcomes is to reduce the variability in our definition of sepsis through the establishment of a consensus definition. Though intended to facilitate study of this important disease in all children, the pediatric consensus definitions for sepsis are not accurate for term neonates and were not created for preterm neonates. Thus, consensus definitions for sepsis in term and preterm infants are needed so that future epidemiologic studies and trials of diagnostic and therapeutic interventions can be interpreted and implemented. Definitions are also needed to help clinicians in determining true sepsis and providing appropriate clinical care. It is therefore critical that a consensus conference be held to establish evidence-based neonate-specific sepsis definitions and to develop a research agenda to validate these definitions.

Abbreviations used

NICU

Neonatal intensive care unit

SIRS

systemic inflammatory response syndrome

ACCP

American College of Chest Physicians

SCCM

Society of Critical Care Medicine

MODS

multiple organ dysfunction syndrome

WBC

white blood cell

EOS

early-onset sepsis

LOS

late-onset sepsis

ANC

absolute neutrophil count

I/T ratio

immature to total neutrophil ratio

CBC

complete blood count

VLBW

very low birth weight

CRP

C-reactive protein

APR

acute phase reactant

CoNS

coagulase-negative Staphylococci

SD

standard deviation

AUC

area under curve